Aims: Prediabetes and diabetes are associated with increased insulin resistance and decreased insulin production, dyslipidemia, and increased cardiovascular disease (CVD) risk. Our goals were to assess lipoprotein subfractions using novel assays in such subjects.Methods: Fasting normal, prediabetic, and diabetic Taiwanese men and women (n = 2,049) had their serum glucose, glycosylated hemoglobin, insulin, total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), HDL3-C, apolipoprotein E-HDL-C, direct low-density lipoprotein cholesterol (LDL-C), small dense LDL-C (sdLDL-C), LDL-TG, and remnant lipoprotein cholesterol (RLP-C) levels measured using novel assays. HDL2-C, LDL-C, and large-buoyant LDL-C (lbLDL-C) were calculated.Results: Prediabetic male and female subjects had significantly higher levels of TG, RLP-C, sdLDL-C, the sdLDL-C/LDL-C ratio, and LDL-TG than normal subjects, and statin treatment abolished this effect in men, but not in women. Diabetic male and female subjects had significantly higher TG and sdLDL-C/LDL-C ratios, and significantly lower levels of HDL-C, HDL2-C, HDL3-C, and apoE HDL-C than normal subjects, as did prediabetic women. Median direct LDL-C levels were > 100 mg/dL in all groups, even in those receiving statin therapy. Calculated LDL-C significantly underestimated direct LDL-C by > 10% in diabetic subjects.Conclusions: Our data indicate that prediabetic subjects were more likely to have significantly elevated RLP-C, sdLDL-C, and LDL-TG, while diabetic subjects were more likely to have significantly decreased HDL-C, HDL2-C, HDL3-C, and apoE HDL-C than normal subjects, and calculated LDL-C significantly underestimated their direct LDL-C. In our view, direct LDL-C and sdLDL-C should be measured and optimized in both diabetic and prediabetic subjects to reduce CVD risk.